search

CVI-SZU / ME-GraphAU

[IJCAI 2022] Learning Multi-dimensional Edge Feature-based AU Relation Graph for Facial Action Unit Recognition, Pytorch code
MIT License
161 stars 38 forks source link

Loss failure to converge #13

Closed LIRENDA621 closed 1 year ago

LIRENDA621 commented 1 year ago

In train stage1, Losses fall very slowly,

lingjivoo commented 1 year ago

Hi, it is normal. Or you can adjust the learning rate.